Facet recommendations from sentiment-bearing content

Abstract

A “Facet Recommender” creates conversational recommendations for facets of particular conversational topics, and optionally for things associated with those facets, from consumer reviews or other social media content. The Facet Recommender applies a machine-learned facet model and optional sentiment-model, to identify facets associated with spans or segments of the content and to determine neutral, positive, or negative consumer sentiment associated with those facets and, optionally, things associated with those facets. These facets are selected by the facet model from a list or set of manually defined or machine-learned facets for particular conversational topic types. The Facet Recommender then generates new conversational utterances (i.e., short neutral, positive or negative suggestions) about particular facets based on the sentiments associated with those facets. In various implementations, utterances are fit to one or more predefined conversational frameworks. Further, responses or suggestions provided as utterances may be personalized to individual users.

What is claimed is: 
     
       1. A computer-implemented process, comprising:
 receiving a machine-learned facet model; 
 the machine-learned facet model automatically generated by applying one or more automated machine-learning processes to a plurality of examples of training data to train the machine-learned facet model; 
 the plurality of examples of training data comprising a plurality of automatically clustered and labeled instances of sentiment vocabulary extracted from sentiment bearing content; 
 applying the machine-learned facet model to evaluate a plurality of samples of sentiment-bearing content to identify conversational topics and facets associated with one or more segments of that content; 
 identifying one or more of the facets that have a consensus based on two or more samples of the sentiment-bearing content; 
 generating a plurality of sentiment-based recommendations about one or more of the identified facets that have a consensus; and 
 applying one or more of the sentiment-based recommendations to change a visual appearance of existing content to indicate a type of sentiment associated with one or more corresponding facets identified in the existing content. 
 
     
     
       2. The computer-implemented process of  claim 1  further comprising:
 generating a plurality of conversational utterances about one or more of the identified facets that have a consensus; and 
 outputting one or more relevant conversational utterances via one or more output devices. 
 
     
     
       3. The computer-implemented process of  claim 2  wherein relevancy of conversational utterances is determined as a reactive response to a user input. 
     
     
       4. The computer-implemented process of  claim 2  wherein relevancy of conversational utterances is determined as a proactive response to current circumstances associated with a user. 
     
     
       5. The computer-implemented process of  claim 2  wherein relevancy of conversational utterances is determined as a response to a user profile associated with a user. 
     
     
       6. The computer-implemented process of  claim 5  further comprising a user interface for modifying of one or more features of the digital avatar. 
     
     
       7. The computer-implemented process of  claim 2  wherein a video of a digital avatar having a plurality of features is used to output one or more of the relevant conversational utterances. 
     
     
       8. The computer-implemented process of  claim 2  further comprising process actions applying a machine-learned translation model to fit one or more of the conversational utterances to one or more predefined personality types. 
     
     
       9. The computer-implemented process of  claim 8  wherein one or more of the predefined personality types is associated with a predefined linguistic style. 
     
     
       10. The computer-implemented process of  claim 8  further comprising a user interface for modifying one or more features of one or more of the predefined personality types. 
     
     
       11. The computer-implemented process of  claim 1  wherein the consensus represents a consensus facet where two or more samples of the sentiment-bearing content refer to the same facet for a particular conversational topic. 
     
     
       12. The computer-implemented process of  claim 1  wherein the consensus represents a consensus sentiment where two or more samples of the sentiment-bearing content are determined as referring to a common sentiment for a particular facet. 
     
     
       13. A system, comprising:
 a general purpose computing device; and 
 a computer program comprising program modules executable by the computing device, wherein the computing device is directed by the program modules of the computer program to:
 receive a machine-learned facet model; 
 the machine-learned facet model automatically generated by applying one or more automated machine-learning processes to a plurality of examples of training data to train the machine-learned facet model; 
 the plurality of examples of training data comprising a plurality of automatically clustered and labeled instances of sentiment vocabulary extracted from sentiment bearing content; 
 apply the machine-learned facet model to sentiment-bearing content to identify and label facets for each conversational topic associated with those facets, and to identify corresponding sentiments; 
 generate a plurality of sentiment-based recommendations about one or more of the identified facets that have a consensus; and 
 generate multiple sentiment-based recommendations that are consistent with the identified corresponding sentiments; 
 select one or more sentiment-based recommendations relevant to a user; and 
 present one or more of the selected sentiment-based recommendations to the user by automatically changing a visual appearance of existing content to indicate a type of sentiment associated with one or more corresponding facets identified in the existing content. 
 
 
     
     
       14. The system of  claim 13  further comprising:
 a program module configured to populate multiple conversational utterances by fitting one or more labelled facets into one or more conversational frameworks that are consistent with the identified corresponding sentiments; and 
 a program module configured to present one or more of the conversational utterances to the user via one or more output devices. 
 
     
     
       15. The system of  claim 14  wherein the relevancy of one or more of the conversational utterances presented to the user is determined in response to a user input. 
     
     
       16. The system of  claim 14  wherein relevancy of one or more of the conversational utterances presented to the user is determined in response to current environmental circumstances associated with the user. 
     
     
       17. The system of  claim 14  wherein one or more of the selected conversational utterances is presented via a combined audio and video display of a digital avatar having a user-selected personality type. 
     
     
       18. A portable computing device comprising:
 a memory configured to store at least one program module; and 
 a processing unit configured to execute the at least one program module to: 
 receive a machine-learned facet model; 
 the machine-learned facet model automatically generated by applying one or more automated machine-learning processes to a plurality of examples of training data to train the machine-learned facet model; 
 the plurality of examples of training data comprising a plurality of automatically clustered and labeled instances of sentiment vocabulary extracted from sentiment bearing content; 
 apply the machine-learned facet model to identify conversational topics and facets associated with one or more segments of a plurality of samples of sentiment-bearing content; 
 determine a consensus for one or more identified facets, the consensus for each facet being based on two or more samples of the sentiment-bearing content; 
 generate a plurality of recommendations about one or more of the identified facets, each of the recommendations being consistent with the corresponding consensus; 
 identify one or more of the recommendations that are relevant to a user; and 
 present one or more of the relevant recommendations by automatically changing a visual appearance of existing content to indicate a type of sentiment associated with one or more corresponding facets identified in the existing content. 
 
     
     
       19. The portable computing device of  claim 18  further comprising:
 generate a plurality of conversational utterances about one or more of the identified facets, each of the conversational utterances being consistent with the corresponding consensus; and 
 wherein one or more of the conversational utterances are fit to one or more predefined linguistic styles. 
 
     
     
       20. The portable computing device of  claim 18  wherein identifying one or more of the recommendations that are relevant to the user further comprises evaluating a user input in combination with a user profile, and in further combination with one or more grounding factors representing current environmental circumstances associated with the user.